• Credit Structure: 3-0-0-0-3

• Introduce diverse learning paradigms, from foundational to emerging approaches.

• Build a conceptual and practical understanding of each paradigm through datasets and algorithms.

• Explore advanced paradigms like meta-learning, multimodal, and generative learning.

• Equip learners with hands-on experience using industry-standard tools and frameworks.

Unit 1: Foundational and Classic Paradigms (? hours)

• • Supervised Learning

    • Supervised learning paradigms

    • Development of ML systems using supervised learning

    • Study and understanding with synthetic datasets

    • Case study:

      • Algorithm: Regression

      • Task: Predicting house prices

      • Hands-on: Implementation using Scikit-learn

  • Unsupervised Learning

    • Unsupervised learning paradigms

    • Development of ML systems using clustering techniques

    • Study and understanding with synthetic datasets

    • Case study:

      • Algorithm: K-Means Clustering

      • Task: Customer segmentation

      • Hands-on: Clustering and visualization using Scikit-learn and Matplotlib

  • Semi-Supervised Learning

    • Semi-supervised learning paradigms

    • Applications in handling partially labeled datasets

    • Study and understanding with synthetic datasets

    • Case study:

      • Algorithm: Label Propagation

      • Task: Text classification with partially labeled data

      • Hands-on: Semi-supervised implementation using Scikit-learn libraries

  • Active Learning

    • Active learning paradigms

    • Importance of data labeling efficiency

    • Study and understanding with synthetic datasets for active annotation

    • Case study:

      • Algorithm: Uncertainty Sampling

      • Task: Medical image classification (active annotation)

      • Hands-on: Active learning pipeline with Scikit-learn

  • Self-Supervised Learning

    • Self-supervised learning paradigms

    • Generating pseudo-labels from unlabeled data

    • Study and understanding of contrastive methods with synthetic datasets

    • Case study:

      • Algorithm: Contrastive Learning (SimCLR)

      • Hands-on: Self-supervised feature learning using

      • Keras, PyTorch Lightning and Hugging Face

Unit 2: Advanced Learning Paradigms (? Hours)

• • Reinforcement Learning

    • Reinforcement learning paradigms

    • Application to decision-making problems

    • Study and understanding with synthetic datasets for basic control tasks

    • Case study:

      • Algorithm: Q-Learning

      • Task: Grid-world navigation

      • Hands-on: Implementation using OpenAI Gym and NumPy

  • Transfer Learning

    • Transfer learning paradigms

    • Adapting pre-trained models for domain-specific tasks

    • Fine-tuning pre-trained models

    • Study and understanding of domain adaptation challenges using synthetic datasets

    • Case study:

      • Algorithm: Fine-tuning ResNet

      • Dataset: Adapting ImageNet features for domain-specific tasks

      • Hands-on: Transfer learning with Keras, PyTorch or TensorFlow

  • Generative Learning

    • Generative learning paradigms

    • Understanding transformer-based approaches

    • Study and understanding of language generation tasks with synthetic datasets

    • Case study:

      • Algorithm: Transformer Network

      • Task: Language modeling

      • Hands-on: Implementing a transformer model using Hugging Face Transformers

Unit 3: Emerging Learning Paradigms (? Hours)

• • Zero-Shot, One-Shot, and Few-Shot Learning

    • Learning with minimal labeled data paradigms

    • Applications to zero/one/few-shot scenarios

    • Study and understanding with synthetic datasets

    • Case study:

      • Algorithm: Siamese Networks

      • Task: text classification, image classification

      • Hands-on: Implementation using PyTorch and Keras

  • Continual Learning

    • Learning without forgetting

    • Incremental learning

    • Life-long learning

    • Case Study:

      • Algorithm: Progressive Neural Networks

      • Hands-on: Develop a model capable of learning a

      • sequence of tasks without forgetting the knowledge gained from previous tasks.

  • Multimodal Learning

    • Multimodal learning paradigms

    • Integration of data from multiple modalities

    • Study and understanding with synthetic datasets for multimodal applications

    • Case study:

      • Algorithm: Cross-Modal Attention Mechanism

      • Dataset: Visual Question Answering (VQA)

      • Hands-on: Multimodal learning pipeline using Transformers and OpenCV

  • Meta-Learning

    • Meta-learning paradigms

    • Applications for rapid adaptation to new tasks

    • Study and understanding of meta-learning techniques with

    • synthetic datasets

    • Case study:

      • Algorithm: MAML (Model-Agnostic Meta-Learning)

      • Application: Rapid adaptation to new tasks

      • Hands-on: MAML implementation with Python libraries like Higher and TorchMeta

  • Other Emerging Paradigms

    • Exploration of cutting-edge learning paradigms in research

    • Case studies and applications discussed based on current trends

• Lecture 1: Introduction

• Lecture 2: Introduction

• Lecture 3: Supervised Learning

• Lecture 4: Supervised Learning

• Lecture 5: Unsupervised Learning

• Lecture 6: Unsupervised Learning

• Lecture 7: Semi-Supervised Learning

• Lecture 8: Semi-Supervised Learning

• Lecture 9: Active Learning

• Lecture 10: Active Learning

• Lecture 11: Self-Supervised Learning

• Lecture 12: Self-Supervised Learning

• Lecture 13: Reinforcement Learning

• Lecture 14: Reinforcement Learning

• Lecture 15: Transfer Learning

• Lecture 16: Transfer Learning

• Lecture 17: Generative Learning

• Lecture 18: Generative Learning

• Lecture 19: Zero-Shot, One-Shot, and Few-Shot Learning

• Lecture 20: Zero-Shot, One-Shot, and Few-Shot Learning

• Lecture 21: Continual Learning

• Lecture 22: Continual Learning

• Lecture 23: Multimodal Learning

• Lecture 24: Multimodal Learning

• Lecture 25: Meta-Learning

• Lecture 26: Meta-Learning

• Lecture 27: Other Emerging Paradigms

• Lecture 28: Other Emerging Paradigms

• 2 Theoretical Assignments (14%)

• 24 Quizzes (36%)

• Project (10%)

• Classroom Notes (10%)

• Activeness in Classes (10%)

• Attendance (10%)

• X-Factor (10%)